Help with "hot mic"

Post your questions & get help from friendly LibriVoxers
Joyfull
Posts: 1129
Joined: March 19th, 2015, 2:42 pm
Location: Lone Star
Contact:

Post by Joyfull » December 4th, 2018, 10:21 am

I am working on a project which requires mixing 2 voices. One of the files has what I've heard as "hot mic" (too loud - even when the volume is low, it sounds too loud in the ear) Is there a way to fix this? Thanks! :D

TriciaG
LibriVox Admin Team
Posts: 40050
Joined: June 15th, 2008, 10:30 pm
Location: Toronto, ON (but Minnesotan to age 32)

Post by TriciaG » December 4th, 2018, 11:34 am

Not that I'm aware of. Yes, you can reduce the volume, but it still sounds slightly distorted, aka "hot".

philchenevert
LibriVox Admin Team
Posts: 16021
Joined: October 17th, 2010, 9:23 pm
Location: By De Bayou
Contact:

Post by philchenevert » December 4th, 2018, 11:55 am

Joyfull wrote:
December 4th, 2018, 10:21 am
I am working on a project which requires mixing 2 voices. One of the files has what I've heard as "hot mic" (too loud - even when the volume is low, it sounds too loud in the ear) Is there a way to fix this? Thanks! :D
Please explain more about mixing 2 voices. What exactly are your trying to do there? In regard to the 'hot mic' I would like to see the wave form but as a first guess, compression may help.
Phil Chenevert, The LibriVox Video Guy
Life is like a game of chess. I don't know how to play chess. .

Need Help? Lots of Helpful Videos Here

tovarisch
Posts: 2562
Joined: February 24th, 2013, 7:14 am
Location: New Hampshire, USA

Post by tovarisch » December 4th, 2018, 12:06 pm

philchenevert wrote:
December 4th, 2018, 11:55 am
In regard to the 'hot mic' I would like to see the wave form but as a first guess, compression may help.
Agree with needing to see the wave, disagree with the guess. Compression makes most of the sound to be perceived louder, so if it's already louder than expected, it's likely been compressed already. Compressing it further will not help.

It's quite possible (and that's where seeing the wave form would help) that the audio has been clipped. Nothing can fix that, unfortunately.
tovarisch
  • reality prompts me to scale down my reading, sorry to say
    to PLers: do correct my pronunciation please

TriciaG
LibriVox Admin Team
Posts: 40050
Joined: June 15th, 2008, 10:30 pm
Location: Toronto, ON (but Minnesotan to age 32)

Post by TriciaG » December 4th, 2018, 12:53 pm

Maybe you could compress the quieter voice, to bring it up to more match the louder one?

DaleInTexas
Posts: 144
Joined: August 15th, 2012, 12:43 pm
Location: West of Cowtown, Texas
Contact:

Post by DaleInTexas » December 4th, 2018, 7:49 pm

Assuming you are using Audacity, have you tried pulling down the track's gain slider down by 10 or 12dB, or whatever gets it sounding about the same level as the lower-volumed track, and bounce it to a new track? De-amplify the offending track. However as Tov said, if it is clipped, it's still going to sound distorted.
Dale
Dale Latham
Down on the Poor Farm
https://twitter.com/dalelatham

Joyfull
Posts: 1129
Joined: March 19th, 2015, 2:42 pm
Location: Lone Star
Contact:

Post by Joyfull » December 5th, 2018, 6:52 am

Thank you all for your replies! You are amazing!

The two files are supposed to be at the same sound level (22 in Audacity), both were originally compressed. When I mixed them, the one with my voice sounds much louder than the other, especially at the beginning of the first word in each sentence, kind of hurts the ear :oops: (even though they have the same volume). What I wanted to know is if there is a way that I can fix my file by itself before I mix them, so I don't have to do each sentence manually.

Maybe you can give me some tips so I record it smoother in the future :D

tovarisch
Posts: 2562
Joined: February 24th, 2013, 7:14 am
Location: New Hampshire, USA

Post by tovarisch » December 5th, 2018, 10:43 am

We would love to give you tips. The problem I have is simple: I am not sure what to suggest because I have not yet figured out what you mean. What's "22 in Audacity"? How do you know they "have the same volume", did you use Checker? Or did you measure it some other way? Or, did you just listen to them separately?

Could you upload two (or three) fragments -- from your recording, from the other, and maybe your preliminary result, then we could listen to them and form our own opinion... Use 'tests-tests' for "MC" in the uploader.
tovarisch
  • reality prompts me to scale down my reading, sorry to say
    to PLers: do correct my pronunciation please

commonsparrow3
Posts: 2409
Joined: January 17th, 2013, 9:16 pm
Location: Rochester, NY

Post by commonsparrow3 » December 5th, 2018, 10:54 am

Joyfull wrote:
December 5th, 2018, 6:52 am
The two files are supposed to be at the same sound level (22 in Audacity), both were originally compressed. When I mixed them, the one with my voice sounds much louder than the other, especially at the beginning of the first word in each sentence, kind of hurts the ear :oops: (even though they have the same volume). What I wanted to know is if there is a way that I can fix my file by itself before I mix them, so I don't have to do each sentence manually.
What you are describing sounds similar to something I've experienced. I've edited together a group song for the LV anniversary the last few years. I'm mixing multiple voices, and I want every singer's voice to be heard. Before I begin, I try to set each track at the same volume. But once I being mixing, there are some voices that dominate and others that fade into the background. They are all the same volume in terms of numbers on a scale, but some voices are simply more resonant, and others are lighter in tone. I have to listen through a few times, before I combine the tracks, to identify which voices need to be brought up and which need to be lowered a bit. It's not something that can be done by any formula of numbers, but has to be done simply by ear. I just keep tinkering with volume levels for the various tracks until my ear says they all sound even.
I've also run into this when editing Dramatic Works. The numbers say they're all at the same volume, but when I put the two speakers lines next to each other in conversation, one seems to be booming away and the other whispering -- again, because one speaker has a lighter voice and the other a more resonant voice. This isn't something the volume levels on the individual tracks would reveal ahead of time -- the numbers say they're the same volume. Only when I begin editing them next to each other does it become noticeable. Usually it becomes noticeable after the first few edits, and then I go back and tinker with the volumes of each file before I finish the editing. Otherwise, I'd have to go through the edited file finding each speaker's lines. And, as you pointed out, doing each sentence manually is a bit of a pain.
I don't know whether what you're trying to do is similar to either of these scenarios, but if it is, maybe my experiences might be relevant.

Joyfull
Posts: 1129
Joined: March 19th, 2015, 2:42 pm
Location: Lone Star
Contact:

Post by Joyfull » December 5th, 2018, 12:15 pm

Thanks Maria, yes, I'm working on a book which contains English and Spanish speakers - For what you said, I guess there is no easy way around it :| . I'll try to lower my volume some more :D

Joyfull
Posts: 1129
Joined: March 19th, 2015, 2:42 pm
Location: Lone Star
Contact:

Post by Joyfull » December 5th, 2018, 12:20 pm

Tovarish,
In Audacity you can check the volume by Highlighting the portion you want to check, then go to Analyze - Contrast (the top tells you the volume of the speaking parts - the bottom of the empty spaces). 22 is ACX preferred volume. I'll upload a portion of the file. :D Thanks!

Joyfull
Posts: 1129
Joined: March 19th, 2015, 2:42 pm
Location: Lone Star
Contact:

Post by Joyfull » December 5th, 2018, 12:40 pm

This is the file I'm talking about. As you can notice, my voice is way fuller - but it seems to hurt my ears when I listen to it.

https://librivox.org/uploads/tests/sample_for_LV-.mp3

Thanks again :D

commonsparrow3
Posts: 2409
Joined: January 17th, 2013, 9:16 pm
Location: Rochester, NY

Post by commonsparrow3 » December 5th, 2018, 3:14 pm

Joyfull wrote:
December 5th, 2018, 12:20 pm
In Audacity you can check the volume by Highlighting the portion you want to check, then go to Analyze - Contrast (the top tells you the volume of the speaking parts - the bottom of the empty spaces). 22 is ACX preferred volume. I'll upload a portion of the file. :D Thanks!
Joyfull wrote:
December 5th, 2018, 12:40 pm
This is the file I'm talking about. As you can notice, my voice is way fuller - but it seems to hurt my ears when I listen to it.

https://librivox.org/uploads/tests/sample_for_LV-.mp3

Thanks again :D
Joyfull --
I've taken a look at your sample file. I've never used the method you describe (Analyze - Contrast), never even realized it existed. I've always used the LV Checker for volume checks. Using that method, I find that your sample file is just a bit on the loud side, slightly over the 92 db which is the loudest end of the LV standard. I tried editing some of the Spanish and English bits into separate files and comparing volumes, and they are both just about that 92 db range, not much difference. But my ear does sense the Spanish voice as "louder", just because it is more resonant. If I were editing these two voices together into one file, I think I'd reduce the volume of the Spanish file by a decibel or so before I start editing them together, and see if that makes them seem more even.

The other thing I noticed was that the Spanish parts do have a lot of clipping. If you look at the wave shape in Audacity, you'll see a lot of flat tops to the waves in the Spanish file, as compared to the pointed or rounded tops of the wave in the English file. This means that when the Spanish file was originally recorded, some parts were so loud that they were outside the space available for them, and the tops were cut off. When the file was compressed, the loud parts were reduced, but the parts that had been cut off couldn't be restored. Those flat tops show where something was cut off. That harshness that "hurts your ears" is probably due to the clipping.

I've been told that if clipping is really bad, there's not much to be done about it, but that if it's not too bad, there are some things that can be done to make it sound better, though not perfect. Unfortunately, I'm not a lot of help with the technical side of things. I only know enough to recognize the problem, but not enough to know how to fix it. This is where someone with more technical know-how like tovarish or Dale or Phil could advise you better. Maybe they could walk you through a way to fix these clipped bits.

Meanwhile, going forward, I'd suggest keeping one eye on your Audacity wave as you record, so you'll notice when you are clipping. When you see it happening, pause and try backing off just a bit from your mike. There's a setting in Audacity that I use to help me be aware of when I'm clipping. Go to 'Tracks' and make sure 'Show Clipping' is checked, and then the blue wave will turn red whenever you are clipping, so your eye will catch it more easily.

TriciaG
LibriVox Admin Team
Posts: 40050
Joined: June 15th, 2008, 10:30 pm
Location: Toronto, ON (but Minnesotan to age 32)

Post by TriciaG » December 5th, 2018, 3:22 pm

Actually, not all flat wave forms were clipped. Sometimes they've got a "hard limiter" of some sort or simply are really, really compressed.

Yes, wave forms that look like a lawnmower went over them, no matter how they got that way, will sound louder than ones where the grass was allowed to grow without much cutting. ;)

GrayHouse
Posts: 452
Joined: October 6th, 2012, 3:27 pm

Post by GrayHouse » December 5th, 2018, 5:03 pm

I agree with Tricia - I think this is agressively compressed/limited rather than clipped. Either way it's unpleasant to listen to.

Obviously there's no magic fix. This is just my quick attempt at a fix:
- https://librivox.org/uploads/tests/Joyfull_-_sample_repair.mp3

I used 2 dynamics tools to repair the sound and restore some dynamic range. Obviously you only want to do it on the problem audio (not the english sections) so it depends how much audio you've got to process as to whether it's a viable solution.

-Ian

Post Reply